Research on a practical de-noising method and the characterization of partial discharge UHF signals

Experimental studies on partial discharges (PD) occurred in oil-insulated devices using ultra-high frequency (UHF) sensors to detect PD signals and using wavelet analysis to process the acquired UHF signals are presented, and a threshold selecting method adapting to various noise circumstances when de-noising UHF signals using wavelet transform is proposed in this paper. The threshold method was based on numerical fitting of standard deviations and manual setting thresholds for brush-fire samples from a large scale. Horizontal comparisons between different geometries and vertical comparisons between different intensities of partial discharges in oil were also discussed and the mathematical tool was wavelet analysis. Multi-resolution analysis (MRA) results show the significant differences that can be used to recognize the discharge pattern and to judge the intensity. Particularly attention was given to needle-plate discharge geometry, which was often used by researchers focusing on the mechanism of pre-breakdown phenomena in liquid dielectrics. The results of needle-plate model show good agreement with electronic theory of liquid pre-breakdown mechanism.